Kimi K2.5/K2.6 1T — B200 vs B300
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and B300 (NVIDIA Blackwell) on Kimi K2.5/K2.6 1T. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
B200 / B300 on Kimi K2.5/K2.6 1T at 56 tok/s/user: 358 / 422 tok/s/GPU, $1.52 / $1.54 per million tokens. B200 is 2% cheaper per token; B300 delivers 18% more tok/s/GPU.
Around the middle of the 37–115 tok/s/user interactivity band, at 76 tok/s/user on Kimi K2.5/K2.6 1T: B200 runs 229 tok/s/GPU at $2.34/M tokens, B300 runs 291 at $2.23/M. B300 is 5% cheaper per token; B300 delivers 27% more tok/s/GPU.
Setting 96 tok/s/user as the target on Kimi K2.5/K2.6 1T, B200 produces 159 tok/s/GPU ($3.41 per million tokens) and B300 produces 201 ($3.25). B300 is 5% cheaper per token; B300 delivers 26% more tok/s/GPU. (Numbers reflect the default 1k/1k · int4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:358.0B300:422.2 | B200:229.2B300:291.4 | B200:159.1B300:200.9 |
| Cost ($/M tok) | B200:$1.515B300:$1.538 | B200:$2.344B300:$2.231 | B200:$3.408B300:$3.247 |
| tok/s/MW | B200:164973B300:194561 | B200:105609B300:134275 | B200:73312B300:92565 |
| Concurrency | B200:~26B300:~32 | B200:~12B300:~8 | B200:~7B300:~4 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.